Resistor availability frustration

[McMonster]

I've finished the design of my latest project which is a digitally controlled dummy load (seems like everyone is building one) and this is the first time I decided to put some thought into building something with specified accuracy and resolution for reading current and voltage levels of the load. So I carefully calculated everything, experimented with several options and came out with something that gives out nice round numbers, specified errors and ranges I'd want. I even optimized it to use only two values of 1% resistors for 1/8 and 8x gains.

But I did a bad mistake of not checking local electronics shops if they carry metalized 1% 0.6 W resistors I wanted to use for dividers and op amp gain settings. I knew most of them carried those, but none has the values that fit my project. It'd be a pain in the ass to redesign and recalculate everything now and I'm looking for options. The best thing I can think of is manually sorting and matching standard 5% carbon resistors using a current source and my 0.03% DC accuracy rated meter. The temperature coefficient would be higher, but I could replace them later when I get a stock of the right resistors somewhere.

Is there any other clever workaround for this? I can just order the damned resistors online, but I've only found a single supplier that actually has any of the matching pair of values in stock but then the shipping cost would be something around 70% of the total cost, even if I order a larger quantity of them.

Also any hints to avoid this situation in the future (beyond "check the suppliers") are welcome.

[free_electron]

that's why in professional meters they don't give a rats ass about the absolute values. as long as tempco is known and the resistors are long term stable they are good. the rest is calibration....

[Jay_Diddy_B]

McMonster,

Can you post this part of your schematic? May be someone can help you.

Do you know what values of resistors you can get? Are they E12, E24 series or random values?

Jay_Diddy_B

[McMonster]

Do you know what values of resistors you can get? Are they E12, E24 series or random values?

I used this calculator, it searches for E24 and E96 values. By the way, is there a calculator that can search for resistor network/divider combination values for arbtirary number of resistors from standard E series and within a defined error?

I'll post entire schematic later after a small rework and cleanup.

[cwalex]

rs components offers free shipping atm in australia. do they have the same deal for your country?

[IanB]

What resistor values do you want and what resistor values can you obtain? Since you don't give resistor values it is hard for anyone to help you.

Furthermore, do you need exact values or do you need ratios? Very often ratio dividers are not critical about absolute values as long as they are within a reasonable range.

[McMonster]

rs components offers free shipping atm in australia. do they have the same deal for your country?

Nope, their shipping is bloody expensive.

[Shuggsy]

If space isn't an issue (is it? do you have a PCB?) then maybe you can stick a few available resistors in series/parallel to get he value you need? It's a pain, but could save you some re-design I suppose. What value are you trying to achieve?

[McMonster]

If space isn't an issue (is it? do you have a PCB?) then maybe you can stick a few available resistors in series/parallel to get he value you need? It's a pain, but could save you some re-design I suppose. What value are you trying to achieve?

The space isn't a big issue judging from previous revision, that's why I asked about a better resistor calculator. Just trying to get a 1/8 divider. I'm querying local shops for the complete lists of values they have or if they source the correct values for me.

[Jay_Diddy_B]

In the E24 series of resistors

91K and 13K will give you a division ration 8 to 1.

You can also use 9.1K and 1.3K


You can also use:

33K and 4.7K will give you a division ratio of 8.02 : 1 which is a 0.25% error


Jay_Diddy_B

[saturation]

that's why in professional meters they don't give a rats ass about the absolute values. as long as tempco is known and the resistors are long term stable they are good. the rest is calibration....

its not a good idea to base a design on a fixed parameter of a component.  In this case, an issue isn't resistor tolerance, but the stability over time and enviromental conditions, mostly heat but others play too like humidity, but not the tolerance.  Its just so happens that low tolerance components also have good stability, but you should base the design on the actual stablity specifications then calibrate out the differences, as freee- suggested.

So I carefully calculated everything, ...I even optimized it to use only two values of 1% resistors for 1/8 and 8x gains.

But I did a bad mistake of not checking local electronics shops if they carry metalized 1% 0.6 W resistors ....

Is there any other clever workaround for this? .......

Also any hints to avoid this situation in the future (beyond "check the suppliers") are welcome.

[McMonster]

that's why in professional meters they don't give a rats ass about the absolute values. as long as tempco is known and the resistors are long term stable they are good. the rest is calibration....

its not a good idea to base a design on a fixed parameter of a component.  In this case, an issue isn't resistor tolerance, but the stability over time and enviromental conditions, mostly heat but others play too like humidity, but not the tolerance.  Its just so happens that low tolerance components also have good stability, but you should base the design on the actual stablity specifications then calibrate out the differences, as freee- suggested.

Good point. I've compared some datasheets and metal film resistors I want to use have an order of magnitude better stability than typical carbon resistors. And they're still dirt cheap, so I think this was a good choice if not for this annoying availability problem. I planned to do calibration in software anyway, with the exception of the pot for LM723 (low drift and temperature compensated reference).

Here's the schematic if someone is interested, I think this will be the final version:

[free_electron]

i hope you have breadboarded that analog part.... it;s going to oscillate like hell.

you need to provide some 'taming' (dampening) across u3B...

[McMonster]

i hope you have breadboarded that analog part.... it;s going to oscillate like hell.

you need to provide some 'taming' (dampening) across u3B...

It was breadboarded two revisions ago and worked fine, I'll recheck tomorrow.

[Jay_Diddy_B]

Hi,

Check Q2 and Q4 on the schematic. I think that Q2 should be NPN and Q4 should be PNP. The way it is drawn on your schematic both transistors are ON if the output of U3B is between 0.5V and 11.5V

Study this thread for ideas on the analog portion of the load: https://www.eevblog.com/forum/projects/did-i-forget-anything-constant-current-load/

Regards,

Jay_Diddy_B

[McMonster]

Check Q2 and Q4 on the schematic. I think that Q2 should be NPN and Q4 should be PNP. The way it is drawn on your schematic both transistors are ON if the output of U3B is between 0.5V and 11.5V

You're right, fixed.

Study this thread for ideas on the analog portion of the load: https://www.eevblog.com/forum/projects/did-i-forget-anything-constant-current-load/

I've seen this thread already, thanks.

What I'm going to do now is check the stability and order the damned resistors online along with the parts for my next project to save money on shipping, I figured out I won't have time to etch and build it this year anyway as I'll be visiting my family. Thank you all for your input.

[McMonster]

I breadboarded the analog part and the oscillations are there, up to 30 kHz. What would you recommend to get rid of them? I tried filtering in the control loop between U3A and U3B, but this causes large settling times and I would not be able to do fast changing or pulse loads.

Everything else seems to work fine by the way, except for that I used standard tolerance components and the gains and levels are totally screwed. And I feel like LM723's datasheet is lying to me with , but I'll investigate this another time.

And I think I should've put this thread in projects section, any mods around?

[JoannaK]

If this were mine (and a hobby) project, I'd just use whatever closest resistors I could get and did calibration of output I and V with the microcontroller software while testing the circuit. It's a lot easier to change few values from code than hunt absolutely optimal (=correct) parts. 

Of course, there may be some accuracy loss, if the Op-amp gain/bias settings ain't just optimal for AD/DA conversions, but those can be fixed at next revision if you'll indeed need the maximum accuracy.

And, wellcome the wonderfull world of electronics. Where lead times for even most mandatory parts can be 40 weeks or more.

[Jay_Diddy_B]

Hi,
Try a resistor of about 2.2K Ohms between the base and emitters of the driver transistors Q2 and Q4. With out this resistor the op-amp has to move 2x Vbe to before the gate voltage changes.

You could also post an updated schematic.

Jay_Diddy_B

[McMonster]

I did some experimenting and reading and this is updated circuit, there are no oscillations over what current range I can get (all the resistance in the wires with the crocodile clips add up to over an ohm), at least on the breadboard. I'll have to do some more reading on dealing with opamps in the future.

If this were mine (and a hobby) project, I'd just use whatever closest resistors I could get and did calibration of output I and V with the microcontroller software while testing the circuit. It's a lot easier to change few values from code than hunt absolutely optimal (=correct) parts. 

I think I'll put some more pots in there for the gains, but 1% stay, I've found the supplier that has them in stock.

Of course, there may be some accuracy loss, if the Op-amp gain/bias settings ain't just optimal for AD/DA conversions, but those can be fixed at next revision if you'll indeed need the maximum accuracy.

This is an experiment in analog/semi-precision stuff, I mostly played with digital electronics before.

And, wellcome the wonderfull world of electronics. Where lead times for even most mandatory parts can be 40 weeks or more.

I come from the boring world of software developement, where the bandwidth is the limit! And this annoys me like hell.

Hi,
Try a resistor of about 2.2K Ohms between the base and emitters of the driver transistors Q2 and Q4. With out this resistor the op-amp has to move 2x Vbe to before the gate voltage changes.

Tried it and no noticeable change in oscillations, just limited the maximum current of the sink.